Project Objective:

Part I consists of building a model using neural networks as a regressor to solve industry based problem.

Part II consists of building a model using neural networks as a classifier to solve industry based problem.

Observations

Univariate Analysis

Outliers

Bivariate Analysis

REGRESSION MODEL BUILDING

Train Loss: 0.278

Test Loss: 0.421

Model is overfitted as test loss is twice of train loss, which has to be less than or equal to train loss

From grapgh it is clear that with increasing epochs the train loss decreases and test loss slightly incecreases

HyperTuning

After getting all best parameters, building the model with all best parameters

Train Loss: 0.364

Test Loss: 0.3705

Model is trained moderately as test loss and train loss is almost equal

But the losses should be as minimum as possible

Tuning parameters manually with using ramdom values to get minimum loss

Train Loss: 0.2415

Test Loss: 0.2353

Model is trained almost very precisely, as both losses are very less and even test loss is less than train loss

Pickeling the best model

#2. Classification MODEL BUILDING

Train loss: 0.2963

Test loss: 1.30

Train accuracy: 0.921

Test accuracy: 0.577

Model is not trained properly as test accuracy is very less than train accuracy and even test loss is very large than train loss

Train Loss: 0.856

Test loss: 1.0886

Train Accuracy: 0.6728

Test accuracy: 0.6029

Model is not trained properly as test accuracy is less than train accuracy and even test loss is very large than train loss

Tuning parameters manually with using ramdom values to get minimum loss and high accuracy

Train Loss: .4142

Test loss: .122

Train Accuracy: 0.844

Test accuracy: 0.9705

Model is trained almost perfectly as test accuracy is more than train accuracy and even test loss is very less than train loss

Pickeling the model

Part 3

Developing a clickable GUI